Detrimental task execution patterns in mainstream OpenMP runtimes

Weinzierl, Tobias; Tuft, Adam; Klemm, Michael

Abstract

The OpenMP API offers both task-based and data-parallel concepts to scientific computing. While it provides descriptive and prescriptive annotations, it is in many places deliberately unspecific how to implement its annotations. As the predominant OpenMP implementations share design rationales, they introduce "quasi-standards how certain annotations behave. By means of a task-based astrophysical simulation code, we highlight situations where this "quasi-standard" reference behaviour introduces performance flaws. Therefore, we propose prescriptive clauses to constrain the OpenMP implementations. Simulated task traces uncover the clauses' potential, while a discussion of their realization highlights that they would manifest in rather incremental changes to any OpenMP runtime supporting task priorities.

Citation

Weinzierl, T., Tuft, A., & Klemm, M. (2024, September). Detrimental task execution patterns in mainstream OpenMP runtimes. Presented at IWOMP 2024, Perth, Australia

• Presentation Conference Type: Conference Paper (published)
• Conference Name: IWOMP 2024
• Start Date: Sep 23, 2024
• End Date: Sep 27, 2024
• Acceptance Date: Jul 16, 2024
• Online Publication Date: Sep 16, 2024
• Publication Date: Sep 16, 2024
• Deposit Date: Jul 16, 2024
• Publicly Available Date: Sep 20, 2024
• Print ISSN: 0302-9743
• Publisher: Springer
• Peer Reviewed: Peer Reviewed
• Series Title: Lecture Notes in Computer Science
• Series Number: 15195
• Series ISSN: 0302-9743
• Book Title: Advancing OpenMP for Future Accelerators
• DOI: https://doi.org/10.1007/978-3-031-72567-8_14
• Public URL: https://durham-repository.worktribe.com/output/2599509
• Publisher URL: https://link.springer.com/conference/iwomp
• External URL: https://arxiv.org/abs/2406.03077